New Unemployment Model Can Outperform Fed and Professional Forecasters

Near Term Unemployment Rate Forecasts Can Be Improved by up to 30 Percent

SEPTEMBER 13, 2012 -
A new economic model can improve unemployment rate forecasts by up to 30 percent, according to research presented today at the Fall 2012 Conference of the Brookings Papers on Economic Activity (BPEA).

Having an accurate forecast of the unemployment rate is important for policymakers, economists, and the business community because the jobless rate not only provides a strong signal about the state of the business cycle in real time, but also because increases in the unemployment rate have preceded the past three recessions, note authors Regis Barnichon of the Barcelona Graduate School of Economics and Christopher J. Nekarda of the Federal Reserve Board of Governors. 

But forecasting the unemployment rate is a difficult task—especially surrounding economic downturns, they write in “The Ins and Outs of Forecasting Unemployment: Using Labor Force Flows to Forecast the Labor Market.” (PDF)

Barnichon and Nekarda develop a model to forecast unemployment based on labor force flows—the number of workers moving into and out of unemployment each month. Their model outperforms other forecasters and basic time-series models by about 30 percent at horizons of up to three months ahead. The model performs especially well around recessions and turning points, when predicting movements in the unemployment rate is more difficult—and more important. Moreover, the model yields even better predictions when combined with a professional forecast.

Download the paper (PDF) »
Find monthly updates »